Abstract
The intelligence of multiagent systems is known to depend on the communication and observation abilities of its agents. However it is not clear which factor has the greater influence. By following an information-theoretical approach, this study quantifies and analyzes the impact of these two factors on the intelligence of multiagent systems. Using machine intelligence tests, we evaluate and compare the performance of collaborative agents across different communication and observation abilities of measurable entropies. Results show that the effectiveness of multiagent systems with low observation/perception abilities can be significantly improved by using high communication entropies within the agents in the system. We also identify circumstances where these assumptions fail, and analyze the dependency between the studied factors.
Cite
CITATION STYLE
Chmait, N., Dowe, D. L., Green, D. G., & Li, Y. F. (2015). Observation, communication and intelligence in agent-based systems. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9205, pp. 50–59). Springer Verlag. https://doi.org/10.1007/978-3-319-21365-1_6
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
